Feed-forward neural networks and topographic mappings for exploratory data analysis

Lowe, David; Tipping, Michael E.

doi:10.1007/bf01413744

Cited by 98 publications

(47 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This was implemented through a neural network approach known as the supervised Neuroscale algorithm. Neuroscale is a clustering process which performs a dimension-reducing, nonlinear transformation of the original input data (Lowe, 1993;Lowe and Tipping, 1996;Nabney, 2004). The supervised version of this algorithm (Low andTipping, 1996:1997) allows for the inclusion of subjective knowledge, regarding the expected relative differences between the classes, to be included in the training process.…”

Section: Stage (Iii) Classification Of Band-limited Features Based Omentioning

confidence: 99%

“…The parameter α varies from 0 -the original, objective, unsupervised method based solely on the distribution of the original data, to 1 -the completely supervised method that is no longer explicitly dependent on the data distribution. A value of α = 0.5 was chosen since it struck a balance between objective and subjective approaches, hence maintaining the spatial topology of the original data whilst improving the visualisation of the feature (Lowe and Tipping, 1996).…”

Section: Appendix the Itakura-saito Distancementioning

confidence: 99%

See 1 more Smart Citation

Neuronal oscillations in the EEG under varying cognitive load: A comparative study between slow waves and faster oscillations

Demanuele

Broyd

Sonuga‐Barke

et al. 2013

Clinical Neurophysiology

View full text Add to dashboard Cite

the support of our participants, Suzannah Helps (Developmental Brain-Behaviour Laboratory, School of Psychology, University of Southampton, UK) for her contributions to the task design, and Daniel Durstewitz (Central Institute of Mental Health, Mannheim, Germany) for useful comments on the manuscript. We also thank all three anonymous reviewers for their constructive comments. AbstractObjective: This study has been specifically designed to investigate very low frequency neuronal oscillations (VLFO, <0.5 Hz) during resting states and during goal-directed tasks of graded difficulty levels, quantify the changes that the slow waves undergo in these conditions and compare them with those for higher frequency bands (namely delta, theta and alpha).Methods: To this end we developed a multistage signal processing methodology comprising blind source separation coupled with a neural network based feature extraction and classification method.Results: Changes in the amplitude and phase of brain sources estimates in the VLF band between rest and task were enhanced with increased task difficulty, but remained lower than those experienced in higher frequency bands. The slow wave variations were also significantly correlated with task performance measures, and hence with the level of task-directed attention.Conclusions: These findings suggest that besides their prominent sensitivity to external stimulation, VLFOs also contribute to the cortical ongoing background activity which may not be specifically related to task-specific attention and performance.Significance: Our work provides important insight into the association between VLF brain activity and conventional EEG frequency bands, and presents a novel framework for assessing neural activity during various mental conditions and psychiatric states. Highlights• We present a novel approach for exploring the functional relationship between very low frequency EEG oscillations (VLFO, <0.5 Hz) recorded at rest and those following the transition to goaldirected tasks of graded difficulty levels.• We show that slow waves (i) are attenuated but not extinguished following a rest to task transition, and (ii) are sensitive to variations in cognitive load; (iii) the rest-to-task changes in the slow wave band correlate with performance measures and (iv) the changes are lower than those in higher frequency bands, tentatively suggesting the VLFOs' contribution to the cortical ongoing background activity.• We present a new signal processing methodology for quantifying changes in band-limited EEG activity -this method, developed to gain insight into the neurophysiological role of the slow waves, could be explored with respect to altered functioning of brain oscillators in psychiatric and neurobehavioural disorders such as schizophrenia and ADHD. C. Demanuele et al., 20123

show abstract

Section: Stage (Iii) Classification Of Band-limited Features Based Omentioning

confidence: 99%

Section: Appendix the Itakura-saito Distancementioning

confidence: 99%

Neuronal oscillations in the EEG under varying cognitive load: A comparative study between slow waves and faster oscillations

Demanuele

Broyd

Sonuga‐Barke

et al. 2013

Clinical Neurophysiology

View full text Add to dashboard Cite

show abstract

“…Neuroscale [20,19] has also been used for multi-objective visualisation [11,8] -but unlike PCA it provides a non-linear mapping. However, although popular across many application domains, both Neuroscale and PCA are oblivious to whether solutions dominate each other, or are mutually non-dominating in multi-objective populations, or what their Pareto shell is.…”

Section: Pareto Dominancementioning

confidence: 99%

Visualising High-Dimensional Pareto Relationships in Two-Dimensional Scatterplots

Fieldsend

Everson

2013

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper two novel methods for projecting high dimensional data into two dimensions for visualisation are introduced, which aim to limit the loss of dominance and Pareto shell relationships between solutions to multi-objective optimisation problems. It has already been shown that, in general, it is impossible to completely preserve the dominance relationship when mapping from a higher to a lower dimension -however, approaches that attempt this projection with minimal loss of dominance information are useful for a number of reasons. (1) They may represent the data to the user of a multi-objective optimisation problem in an intuitive fashion, (2) they may help provide insights into the relationships between solutions which are not immediately apparent through other visualisation methods, and (3) they may offer a useful visual medium for interactive optimisation. We are concerned here with examining (1) and (2), and developing relatively rapid methods to achieve visualisations, rather than generating an entirely new search/optimisation problem which has to be solved to achieve the visualisation-which may prove infeasible in an interactive environment for real time use. Results are presented on randomly generated data, and the search population of an optimiser as it progresses. Structural insights into the evolution of a set-based optimiser that can be derived from this visualisation are also discussed.

show abstract

“…Kruskal, 1964;see Shepard, 1980 for an overview), have also been applied to data visualization, exploration and analysis (e.g. Mao and Jain, 1995;Lowe and Tipping, 1996).…”

Section: Spatial Visualizationsmentioning

confidence: 99%

Visualizations of binary data: A comparative evaluation

Lee

Butavicius

Reilly

2003

International Journal of Human-Computer Studies

View full text Add to dashboard Cite

Data visualization has the potential to assist humans in analysing and comprehending large volumes of data, and to detect patterns, clusters and outliers that are not obvious using nongraphical forms of presentation. For this reason, data visualizations have an important role to play in a diverse range of applied problems, including data exploration and mining, information retrieval, and intelligence analysis. Unfortunately, while various different approaches are available for data visualization, there have been few rigorous evaluations of their effectiveness. This paper presents the results of three controlled experiments comparing the ability of four different visualization approaches to help people answer meaningful questions for binary data sets. Two of these visualizations, Chernoff faces and star glyphs, represent objects using simple icon-like displays. The other two visualizations use a spatial arrangement of the objects, based on a model of human mental representation, where more similar objects are placed nearer each other. One of these spatial displays uses a common features model of similarity, while the other uses a distinctive features model. The first experiment finds that both glyph visualizations lead to slow, inaccurate answers being given with low confidence, while the faster and more confident answers for spatial visualizations are only accurate when the common features similarity model is used. The second experiment, which considers only the spatial visualizations, supports this finding, with the common features approach again producing more accurate answers. The third experiment measures human performance using the raw data in tabular form, and so allows the usefulness of visualizations in facilitating human performance to be assessed. This experiment confirms that people are faster, more confident and more accurate when an appropriate visualization of the data is made available. r

show abstract

Feed-forward neural networks and topographic mappings for exploratory data analysis

Cited by 98 publications

References 28 publications

Neuronal oscillations in the EEG under varying cognitive load: A comparative study between slow waves and faster oscillations

Neuronal oscillations in the EEG under varying cognitive load: A comparative study between slow waves and faster oscillations

Visualising High-Dimensional Pareto Relationships in Two-Dimensional Scatterplots

Visualizations of binary data: A comparative evaluation

Contact Info

Product

Resources

About